A storage area network (SAN) is storage connected in a fabric, usually through a switch, that enables many different...
servers to easily access the storage. From a server application and OS standpoint, there is no visible difference between accessing data storage in a SAN or storage that is directly connected. SANs, like direct-attached storage (DAS), enable block access to data.
Network-attached storage (NAS) is a method of remote file serving. Rather than using the software on the file system, file access is redirected to another device using a remote protocol, such as SMB or NFS. That device operates as a server with its own file system, handling the file I/O and enabling file sharing and centralized data management.
The SAN vs. NAS decision usually comes down to the type of data you are storing. With block I/O, SAN is used and, with file I/O, NAS. In addition, when comparing SAN vs. NAS, keep in mind that NAS eventually turns the file I/O request into block access for the storage devices attached to it. SANs are also the preferred choice for structured data -- data that resides in a relational database. While NAS can handle structured data, it is usually used for unstructured data, which consists largely of files, email, social media, images, videos, communications and almost any type of data outside of relational databases.
Object I/O for storage has become more prevalent, thanks mainly to its overwhelming use in cloud storage. As a result, the clear divide between SAN being used with block storage and NAS with file storage is becoming blurred.
As vendors move from block or file to object I/O for their storage needs, users still want to access data in the way they are used to: block storage for SAN or file storage for NAS. Vendors are offering systems with front ends that present a NAS or SAN experience, while the back end is based on object storage.
File vs. block vs. object
File I/O storage reads and writes data in the same manner as the user does on a drive on a computer, using a hierarchical structure with files inside folders that can be inside more folders. NAS systems commonly use this approach, and it has a number of benefits:
- When used with NFS and SMB, the most common NAS protocols, a user can copy and paste files or entire folders just like they can on local and external drives.
- The IT department can easily manage these systems.
Block I/O storage treats each file or folder as various blocks of smaller bits of data and distributes multiple copies of each block across the various drives and devices in a SAN system. The benefits of this approach include:
- Greater data reliability. Data can still be accessed if one drive or a number of drives fail.
- Faster access. Files can be reassembled from the blocks closest to the user and don't need to pass through a hierarchy of folders.
Object I/O storage treats each file as a single object, similar to file I/O, and doesn't have a hierarchy of nested folders like block I/O. With object storage, all files or objects are put into a single, enormous data pool or flat database. Files are found based on the metadata that is already associated with the file or added by the object storage OS.
Object storage has been the slowest of the three methods and is mainly used for cloud file storage. But recent advances in the way metadata is accessed and increased use of speedy flash drives have narrowed the speed gap among object, file and block storage.
Using NAS and SAN
The main difference with SAN vs. NAS is reflected in how each type of storage appears to the user.
A NAS system or device is attached to a network via a standard Ethernet connection, so it appears to users like any other network-connected device. The user connects to the NAS to access files on it. The NAS device has an OS that manages the writing and reading of any data the user's computer requests.
Once it has been mounted on a user's computer, a SAN will appear as a local drive. That means it will function as a local drive, and the OS on the user's computer will handle the commands to read or write data. This enables the user to treat it like any other local drive, including the ability to install software onto it.
SAN vs. NAS connections
A NAS system can be one server or a set of drives or servers in a single device. This gives the NAS system a direct connection to the network, generally using an Ethernet cable connected to an Ethernet switch.
In contrast, a SAN is a pool of drives, devices or servers, connected together by a network fabric, such as iSCSI or Fibre Channel (FC).
Ethernet and fabric have competed on the basis of speed for years. However, the advantage has always gone to fabric because it has a more direct connection that doesn't have to go through the TCP/IP handling of an Ethernet connection. Given that, when data speeds are equal, fabric ends up with the I/O speed advantage because the data gets touched less often as it travels between storage and the user.
Ease of use is a key advantage of NAS. Metadata in a NAS system is hierarchical and readable. Users can use a simple file system browser to see file names and organize them into easily named folders.
With NAS, users can collaborate and share data, regardless of where they're located. NAS makes it easy to access files and folders from any network-connected device.
NAS also provides high capacity at a lower cost than SAN. NAS devices consolidate storage in one place and support data management and protection tasks, such as archiving, backup and cloud storage. And NAS can handle unstructured data, such as audio, video, websites, text files and Microsoft Office documents.
NAS appliances can be outfitted with more or larger disks to expand storage capacity. This approach is referred to as scale-up NAS. They also can be clustered together for scale-out storage. Higher-end NAS devices can accommodate enough disks to support RAID.
NAS enables Portable OS Interface-compliant file access, facilitating centrally managed security and file access and ensuring that multiple applications can share a scale-out NAS device without one application overwriting a file that another application is using.
NAS is not fast enough to meet the needs of high-performance applications. It can get slowed down even further if too many users overwhelm a system with simultaneous requests. The use of flash storage in newer NAS systems, either in conjunction with HDDs or as an all-flash system, alleviates the speed problem, however.
Scalability issues can arise with NAS. Adding too many NAS devices can lead to NAS sprawl, especially if all the devices must be managed separately. Clustered, or scale-out, NAS was devised to mitigate that problem.
Data integrity can become an issue, because file systems store metadata and file content across a logical or physical disk volume. If the file server loses power, the system must perform a file system check, also called a fsck, to validate the state of the data. The delay involved with doing a fsck can be significant depending on the NAS system.
NAS' use of RAID can also be problematic because RAID is reaching scalability limits. Rebuild times can take days, a situation that will only get worse as multiterabyte capacity drives become more common.
SANs treat raw storage as a pool of resources that IT can centrally manage and allocate when it's needed. Because SAN connects over the network fabric, data transfers and access times using SAN are faster than NAS, all things being equal.
SAN systems are highly scalable; capacity can be added as required. Other reasons for deploying SANs include continuous availability and resilience. Highly available SANs are designed to have no single point of failure, starting with highly available SAN disk arrays and switches with redundant critical components and redundant connections to the SAN.
Cost and complexity are the main disadvantages of SANs. The hardware for these systems is expensive, and building and managing them require specialized knowledge and skills.
SAN is far more complex than NAS, with dedicated cabling -- usually FC, but Ethernet can be used -- as well as dedicated switches and storage hardware. FC was developed specifically for storage because Ethernet was not reliable enough to transmit block data before advances were made to the protocol over the past decade. But FC SANs require specialized expertise, as well as dedicated connectivity.
And while SANs are highly scalable, scaling a SAN array vertically is limited. Once the scale-up limit is reached, it's necessary to move to a higher-performance array or to add multiple arrays. An increasing number of SAN disk arrays are avoiding this problem by supporting horizontal scale-out where storage nodes are added that scale capacity and performance simultaneously.
How DAS fits in
DAS is a dedicated server or storage device that is not connected to a network. The simplest example of DAS is a computer's hard drive. To access files on DAS, a user must have access to the physical storage.
DAS can outperform NAS, particularly for compute-intensive software programs. However, with DAS, the storage on each device must be managed separately, adding a layer of complexity to the management of the system. DAS systems generally don't offer advanced storage management features, such as replication, snapshots and thin provisioning, that are common in SAN and NAS.
DAS also does not enable shared storage among multiple users. And because only one host accesses a DAS device, only a portion of the available storage ends up being used.
The rise of unified storage
The emergence of unified storage has given storage administrators the flexibility to run block or file storage on the same array. These multiprotocol systems consolidate SAN block-based data and NAS file-based data on one storage platform. Customers can start with either SAN or NAS and add support and the appropriate connectivity later. Or they can buy a storage array that supports SAN and NAS.
Pros and cons of unified storage
Computer Weekly Editor Antony Adshead talks with Andrew White, former GlassHouse Technologies consultant (now advisory engineer at Dell EMC), about unified storage's challenges and benefits.
Unified storage uses both file and block protocols. It can use file protocols, such as SMB and NFS, along with block protocols, such as FC and iSCSI.
One advantage of these systems is that they require less hardware than traditional storage systems. And newer unified storage offerings are incorporating cloud storage and storage virtualization.
Motherboards may have given birth to the future
The greatest amount of action and excitement today comes from extending the nonvolatile memory express (NVMe) protocol over fabric.
The NVMe protocol is the fastest way to connect a flash storage device directly to a computer's motherboard, communicating via the Peripheral Component Interconnect Express bus. And it greatly outperforms an SSD connected via SATA. Imagine if you could extend that speedy NVMe connection across the fabric that knits together a SAN system.
To be fair, NVMe can't be used to transfer data between the remote end user and the storage array, so a messaging layer must be used. This makes NVMe seem more like an Ethernet-connected NAS system, which uses Ethernet's TCP/IP protocol to handle data movement. But NVMe over Fabrics developers are working on using remote direct memory access (RDMA) to make that messaging layer have the least impact on speed. Among the various types of RDMA being proposed are RDMA over Converged Ethernet, Internet Wide Area RDMA Protocol and InfiniBand, which is in use in high-performance computing systems.
Stand-alone SAN/NAS keeps wary eye on hyper-converged storage products
Object-level storage is ready to replace NAS
Take these precautions when integrating NAS and SAN
Dig Deeper on Primary storage devices
Related Q&A from Randy Kerns
Logical unit numbers are a logical abstraction between a physical disk device and applications. Learn more about LUN use cases and LUN security ... Continue Reading
What is the one hidden gotcha that you'd advise users about if they were shopping for an all-flash storage array? Continue Reading
How much control do you have with all-flash storage arrays? How much control do you have over how arrays handle your data? Do you control the caching? Continue Reading
Have a question for an expert?
Please add a title for your question
Get answers from a TechTarget expert on whatever's puzzling you.